keycap profiles for a flat (non-sculpted) keyboard

[jacobolus]

I was posting about SP's DSS keycap family over in the 'keycaps' subforum http://geekhack.org/index.php?topic=62417, but some of this is probably as or more relevant here in the 'ergonomics' subforum. I’ve been finding – in some wood CNC case/plate experiments – that adding extra height to the further-away rows of keys so that they can be more easily reached makes a nice improvement. (I still don’t have a full keyboard wired up, so take this with a grain of salt.)

Anyway, a sculptured keyboard direct-wired together is obviously not something which can be mass-produced, so it would also be nice to figure out the best we can do given typical PCB/plate construction.

I think with either a combination of DSS/SA(/DSA?), or with just DCS keycaps, it’s possible to make a nicer more ergonomic profile, especially for column-staggered boards like the Ergodox, Atreus, Axios, Keyboard.io, etc.

Something along the lines of:


The key is skipping row 3, typically used for the home row, and shifting all the keycap profiles by one row. As a result, there are nice height jumps when reaching for further-away rows, as well as a nice angle to hold fingers when they reach inward for bottom-row keys.

Has anyone ever experimented with mixing up DCS rows like this (would probably work best with blank keycaps .

[jacobolus]

I said this in the keycaps subforum thread, but might well repeat here:

The hybrid profile is better because there is an increased vertical step between the home row and each row further back. Here’s a diagram showing a bit better.


(note this shows SA row 2 on the number row, rather than the SA row 3 shown above, and a non-flipped DSS row 1 for the Q row; I need to test some of these in person to decide exactly what set works best)

The idea is, as you unbend your finger (extend the two distal joints) to reach further-away rows, the fingertip naturally travels in a direction upward and outward. Therefore, having the further keys be at the same height as the home row keys requires you to reach down by either moving your hand or flexing your first finger joint, the former of which is slow especially if you need to type keys on different rows with the same hand and the latter of which reduces the strength available for pressing the key and I suspect increases fatigue for keys with long travel. (It would be nice to actually have some proper scientific evidence here, unfortunately all the studies I’ve seen on related subjects are pretty crappy.)

Anyway, by raising the height of the further away keys, you make them easier to reach. The closer rows, by contrast, are reached by flexing mainly the second finger joint, and I don’t think they benefit much from being lower. (Though again it would be nice to see some comprehensive empirical testing.)

[jacobolus]

I’d really love to figure out a way to mass-produce a keyboard along the lines of the DataStealth, with a profile like:


cf. http://geekhack.org/index.php?topic=55099.msg1236269#msg1236269


But I think it’s going to be pretty expensive to construct something like that at any kind of scale.

[jacobolus]

[Oobly]

...

The idea is, as you unbend your finger (extend the two distal joints) to reach further-away rows, the fingertip naturally travels in a direction upward and outward. Therefore, having the further keys be at the same height as the home row keys requires you to reach down by either moving your hand or flexing your first finger joint, the former of which is slow especially if you need to type keys on different rows with the same hand and the latter of which reduces the strength available for pressing the key and I suspect increases fatigue for keys with long travel. (It would be nice to actually have some proper scientific evidence here, unfortunately all the studies I’ve seen on related subjects are pretty crappy.)

Anyway, by raising the height of the further away keys, you make them easier to reach. The closer rows, by contrast, are reached by flexing mainly the second finger joint, and I don’t think they benefit much from being lower. (Though again it would be nice to see some comprehensive empirical testing.)

You can do this by changing the angle of the board, raising the back. If you use the centre of the keycaps as a reference, the heights in your profile mockups make a pretty straight line. The only way to really get a good curve to match your finger extension is with curved keywells / backplane or use something like SA R1, DSS R1, DSS R2, SA R4 for the 1, Q, A and Z rows. You have to move your arms anyway to reach the F key row.

I think the main advantage of having curved keywells is that it brings the keycap tops closer together, so you don't have to extend your fingers as much. It also presents the surfaces at a nice angle which feel better.

[jacobolus]

You can do this by changing the angle of the board, raising the back.

No, that’s counterproductive, because it changes the orientation of the switches, forcing you to press outward as well as down. Also, it makes the closer letter row and the modifier/spacebar row too short.

Ideally we want the keyboard at a proper height such that we can have a slight *negative* board angle, with our hands hovering comfortably above.

Anyway, adding some height stagger to the keys actually makes them quite a bit easier to reach, quite similar to the curved keywell.

For me, the ideal is a slightly rectangular keycap, with key spacing like 16mm x 19mm instead of the usual 19mm x 19mm, and a keyboard construction method which raises the height of the further away rows of switches, including a row of switches angled below the letters which has the switches angled somewhat horizontally, and a thumb section with a couple of different angled sections.

However, building a fully three-dimensional keyboard is expensive and difficult to mass-produce, and using non-standard keycap shapes requires some source. So these ideas are something of a best alternative given 1 plate/pcb per hand.

[vvp]

Ideally we want the keyboard at a proper height such that we can have a slight *negative* board angle ...

... and a keyboard construction method which raises the height of the further away rows of switches ...

Aren't you in conflict with yourself? Negative board angle makes further away switches lower and at the same time you want them higher. You should decide what you want

The only way how you can make sense is you actually caring about the exact angle your finger has (when pressing a key) to the switch axe. But at agnles around 5° you have shown at your pictures below, this just does not make sense. You are talking about difference in force ... well probably below 1%.

[jacobolus]

Aren't you in conflict with yourself?

No. To see why requires thinking about / inspecting your hands and the way your fingers move, paying close attention to which joints you flex/extend to locate key tops and which joints you flex to actuate the keys.

At some point maybe I’ll try to make some better photos/videos to explain.

The only way how you can make sense is you actually caring about the exact angle your finger has (when pressing a key) to the switch axe.

Yep, that’s right. I think the switch axis is important.

[vvp]

The only way how you can make sense is you actually caring about the exact angle your finger has (when pressing a key) to the switch axe.

Yep, that’s right. I think the switch axis is important.

In my opinion:

• Only if the difference is big, like significantly more than 5°. Maybe, it can make sense to consider it above 15° (force difference more than ~ 5%), but probably only at about 25° (force difference more than ~ 10%). Try to measure it before you invest in 5° tilt compensated with keycaps. Let us know the result if you do measure it.
• You will not achieve enough difference in the angles only by tilting keyboard and compensating with different keycap types. That is the reason I'm trying to build Katy.
  

[jacobolus]

Only if the difference is big, like significantly more than 5°. Maybe, it can make sense to consider it above 15° (force difference more than ~ 5%), but probably only at about 25° (force difference more than ~ 10%). Try to measure it before you invest in 5° tilt compensated with keycaps. Let us know the result if you do measure it.

The issue is not really amount of force required. It’s about the finger motions needed. I’ve never used a Kinesis Advantage for more than a few hours, but the handful of times I tried one I found the top row keys – and especially the keys in the top corners – incredibly awkward. Their spatial position is fine, but their axis of motion makes pressing them very uncomfortable (for me). Also, I don’t think the Kinesis Advantage properly positions the home row to align to the finger positions of a relaxed hand. Both the relative height (off the table) and the depth (distance away from the body) of various columns is in my opinion quite suboptimal.

That is the reason I'm trying to build Katy.

As far as I can tell you are designing the finger section basically by starting from the Kinesis Advantage, rather than reasoning from first principles about the mechanics of the hand. As a result, I suspect that I would find your design just as awkward/uncomfortable as the Kinesis Advantage, with the one benefit that you can independently move and orient the two hand pieces. I have basically the same criticism of your thumb section: it’s not positioned or oriented starting from a consideration of human hand mechanics.

All the best luck with it though. It’s great to build things and experiment!

By the way, a lot of the problems the Kinesis Advantage has are design constraints due to their bent PCB construction method. I recommend you study the Maltron, which in my opinion is a noticeably better design in several important details (but unfortunately is not amenable to mass production so is an expensive niche product).

[vvp]

It does not look like Maltron uses much the fact that they do not use flex PCBs. Well if angles below 15° are important to users then maybe something can be found there. But to me, it looks the same. There is one significant difference though. Kinesis thumb cluster is much higher (and less tilted). Well and Maltron's bottom row seems to be less tilted which looks like disadvantage to me (but IIRC Sordna would like that).

Really, if I would care so much about angles, I would go more the way Oobly went: discard the keys far away from the home row and compensate by more layer shifting. You cannot get too bad angles when all far away keys are not there. The problem is I'm not a big fan of layer shifting.

[jacobolus]

The Maltron still has some problems I think. In particular I don’t like the way they deal with far-away rows. But they do a better job with the orientation/positioning of columns between different fingers, especially the pinky, compared to the Kinesis Advantage. Also, their thumb section is better positioned and better oriented than the Kinesis thumb section.

I agree: it’s useful to scrap every key you can get away with scrapping. Layer shifting is much better overall than having huge numbers of extra keys. I think it’s possible though to reach two rows of keys beyond the "home row" for each finger, if necessary, and that can be made pretty comfortable as long as the rows have enough height step between them. It’s also somewhat possible to have two rows closer than each home row position, but the nearest keys are pretty hard to reach (unless rotated up 60–90° to afford a squeezing rather than pressing motion).

But more rows aren’t strictly necessary, and a keyboard with only 3 rows of finger keys is very usable.

[jacobolus]

On a handful of keyboards I have with extra-tall f-key row keys, I’ve been testing out row-shifting standard-profile keycaps (i.e. using f key caps for numbers, using number caps for qwert row, using qwert row caps for the home row).

It’s so nice!

The number row is very dramatically easier to reach, while still being very comfortable to press, and the qwert row is also easier than normal. I have basically no problems at all with my fingers running into the corners of keycaps or anything like that, as I worried might be an issue. Having the bottom row keep its original profile works just fine, though I think e.g. DCS has a nicer slope (steeper) than something like Cherry profile, where the zxcvb row is flatter.

Switching between a completely flat profile (DSA), a standard sculptured profile, a curved-backplate Model M profile, and the row-swapped new profile, the advantages of the new profile are quite apparent. And this is on a standard layout keyboard. It should be even better still on a column-staggered board.

I also now understand why on a typewriter the bottom row was considered so hard to type (and therefore why alternative typewriter layouts tried to put uncommon letters down there) – if there’s a big step from row to row on every row including the bottom, then to reach the bottom row requires either an awkward finger reach or some hand movement. By contrast, on a standard keyboard (or especially on a totally flat keyboard like a modern laptop’s) the bottom row is pretty easy to reach but the top rows are a big pain.

I’m kind of amazed that no one ever tried this before, or that it’s not a more common thing, considering what an easy change it is.

[Oobly]

... I think it’s possible though to reach two rows of keys beyond the "home row" for each finger, if necessary...
But more rows aren’t strictly necessary, and a keyboard with only 3 rows of finger keys is very usable.

I agree. On a "normal" keyboard, the different finger lengths and angled hands allows the number row to be fairly easily reached by any finger other than the index finger. This also makes the lower row more difficult to reach for the same fingers. It just highlights again how unergonomic the normal layout is. I have found 3 keys per column in a staggered column layout works very nicely. I like my layers I use symbols on "Fn" and F keys and numbers on "Fn+Shift" layers using left thumb to press both (thumb key row closest to finger keys).

I use OEM blanks on my DIY board, so I'll try using num row profile on Q row and report back here about how it feels.

I will be ordering the Pulse SA set, I'm interested to see how contoured SA feels, too

[Oobly]

Thought I'd post this pic of original Cherry profile here, since it has higher Row F keys than DCS Row 5. Makes me think I should have added the F and A Rows kit to my GMK Dolch set. If you "compress" it to A, B, C, E, F (dropping D to get bigger step from C to E) it looks pretty good. Or A, C, F for 3 row layouts.

[PieterGen]

Great posts and research! I am planning to build one of jacobolus' ideas, namely this one:
. But with some modifications, of course. Two changes were suggested to me by Xah Lee:
- remove the Esc and Backspace and put them in an extra column in the middle (= more hand separation within the same 60% border).
- Remove the lower corner 1 3 7 9 and arrow. Make them into a row instead, flush with the thumb keys row.  I'm not sure about this....

For my build I will definitely use your ideas on key profiles!

What I want is a portable keyboard. Influenced by jacobolus'concepts and Oobly's builds... If that is alright with you guys!  This means:
- small (60% size)
- one-piece, for simplicity
- flattish..... completely flat, or some tenting and the thumb keys on a lower plane... I won't go as far as Oobly, I want it to be flattish... But some tenting and level difference may be beneficial already.
- thumb keys: I'm torn between the Oobly's idea and the Esrille thumb keys.... 


I will use my own layout of course, which I calculated with Michael D1cken's mtgap software (I can't spell his name correctly due to forum software!) , using a corpus that represents my workflow (85% dutch prose, 10% english prose, the rest other languages, and just a wee bit of coding web stuff). It performs (for my use) better than other known layouts and it looks like this: 
. u o p y x c l b v
a i e n h m d r t s
: , ? k q f g w j z

Anyway.... sorry for derailing, I may use Oobly's setup, "flatten" that and put it on a 60% groundplate. Suppose I did,
- what key rows/profiles would be best for the hands, and which for the thumbs?
- if I put it on 60% groundplate, what would you put in the middle? I'd love some dedicated keys for much used functions, such as ctrl-a, ctrl-x, ctrl-v, ctrl-z, ctrl-z etc...
- Question for Oobly: I suppose you have to keep thumb keys pushed down, for the layers to work. Suppose you have to inputs lots of numbers, doesn't that get old? Or do you have a sort of "capslock" function... (omg what did I say     ) you know, press once for the layer to activate, press again to get back to the normal layer...

Finally: for a 3 row setups (like Oobly's), would this work (I'm using qwerty row names now, for simplicity):
DCS row5 for the Q row
DCS row 2 or row 3 for the A row (=home row),
DCS row 4 for the Z row

Or would the steps be too big?
 

Hope all this is sufficiently on topic!

[jacobolus]

Note, the premise of my little board there was to see what kind of “ergonomic” keyboard could be made within the constraints of fitting in a standard 60% case.

The keys in the corner are not supposed to be number keys, but rather arrows / pgup/pgdn/home/end. (Just the arrows and nav cluster keycaps I had from the Cherry boards I harvested keycaps from happened to be the relegendable type, which I don’t like too much, so I used the keycaps from the numpad instead. But while we’re talking keycap profiles, I think the profiles of the numpad keys for arrows and nav cluster work pretty well, when put together like that.

To be honest, if it were up to me I’d skip the dedicated arrows and pgup/pgdn/home/end altogether, and just put those keys on a layer. The main reason they’re there is (1) when cramming an angled keyboard into a predefined rectangle, there’s extra space in the corners, and (2) since I think this keyboard design could appeal to a somewhat broad audience, I wanted to include keys that many people find useful.

If you aren’t going to try to use a standard case and don’t care about a rectangular shape, and you’re wiling to come up with your own keymap, you might want to just skip those keys.

Also, you want to test out the amount of stagger to use between columns. That rectangular board of mine has stagger reduced relative to the optimum amount, to help it fit within the available space.

- remove the Esc and Backspace and put them in an extra column in the middle (= more hand separation within the same 60% border).

Note that backspace in my board would be on one of the 1.5u thumb keys. The blank keys at the right and left are shift keys, and the key in the top right corner is return.

inre tenting, I think the ideal for a keyboard with the two halves attached to each-other and fairly close together is something like 30-45° of tilt. Along with the front raised a bit higher than the back, and the columns angled a bit toward the center.

The Esrille board seems like a good design, but I’d drop some of the far pinky keys on a board that small, and also drop the outside thumb keys (and maybe add another thumb key on each side in the direction of the center), increase the amount of column stagger, dump the F keys (or at least move them), and increase the amount of tenting.

- what key rows/profiles would be best for the hands, and which for the thumbs?
- if I put it on 60% groundplate, what would you put in the middle? I'd love some dedicated keys for much used functions, such as ctrl-a, ctrl-x, ctrl-v, ctrl-z, ctrl-z etc...

I recommend getting a bunch of keycaps and testing it out. It’s hard to come up with a perfect setup in the abstract, without putting your hands on it.

- Question for Oobly: I suppose you have to keep thumb keys pushed down, for the layers to work. Suppose you have to inputs lots of numbers, doesn't that get old?

In practice this is really no big deal. If you need to do data entry or something, add a locking layer. Otherwise, I don’t think it’s worth it.

Finally: for a 3 row setups would this work
DCS row5 for the Q row
DCS row 2 or row 3 for the A row (=home row),
DCS row 4 for the Z row

I think the jump from row 2/3 to row 5 might be slightly severe, but test it out! I suspect that you might prefer DCS row 1, row 2, and row 4 on a 3-row board.

For the pinkies, you might want to use row 5, row 1, and row 4 instead. Having the pinky caps be a bit taller can be an advantage.

Here’s one possible shape:

[jacobolus]

You may also be interested in: http://geekhack.org/index.php?topic=56095

[vvp]

Also, you want to test out the amount of stagger to use between columns. That rectangular board of mine has stagger reduced relative to the optimum amount, to help it fit within the available space.

Interesting, the stagger you already have there looks like a big one to me.

It is also 10 kinds of keycaps. Is some company providing them all?
That may be expensive without a good group buy.
Well not as expensive as a contoured keywell(*). Keycaps can compensate for non-contoured keywell somewhat, so it is a good approach.

(*) If you do not already have a cheap 3dPrinter at home. In such a case a contoured keywell is cheap.

[jacobolus]

Also, you want to test out the amount of stagger to use between columns. That rectangular board of mine has stagger reduced relative to the optimum amount, to help it fit within the available space.

Interesting, the stagger you already have there looks like a big one to me.

We’re talking about this one:

[vvp]

Ok, that makes sense now.

[PieterGen]

Thanks jacobolus. I'll make some cardboard and keycaps mockups, I have some cheapo keycaps for that, I pryed them off a €10 junk keyboard.

[jacobolus]

This is all assuming that we’re dealing with a high desk relative to the chair, so we have a steeply angled keyboard; this is not especially ergonomically friendly, but it’s what basically every typewriter and keyboard did and still does, because most people have really bad workspaces set up. For other desk heights, just tilt everything the appropriate amount.

The second to last picture here is roughly what the Maltron / Kinesis Advantage does (though it’s not super accurate... I just sketched it quickly, rather than basing it on direct measurements). The last picture is what I think a keyboard should be like, if manufacturing costs are no object. Note, I think spherical tops are fine too; the positions/orientations of the switches/keycap tops are what I’m focusing on here.

[jacobolus]

I finally got hold of a bunch of DCS row 5 keycaps (they’re about the same height/tilt as SA row 2/4, if someone wants to simulate).

I really like the (from top to bottom starting at the number row) row 5, row 1, row 2, row 4, row 4 profile, with keys to be pressed by the thumb using flipped row 1 or 2.



In the diagram below, purple = standard DCS, green = the profile from the picture above, black = close to my ideal for a standard-layout keyboard:

[Oobly]

I put some SA caps on my ergo board (which is essentially "flat" front to back) as a test. Row 1, Row 3, Row 4:





Feels really good. When I get more Row 1 caps I will try Row 1, Row 3, Row 1 inverted (since the Row 4 caps feel like they could do with a bit more angle and height). With only 3 rows to worry about it's easier to find a set of angles and heights that works, but the same principal applies.

[yasuo]

finally oobly ur sa

interisting its like Rivertron


i've tried bottom row i increase/tilt forward the switches feel more natural indeed
you can try with plate no pcb need glued for permanent/stabile

[Coreda]

We’re talking about this one:

Show Image

Is this a layout-only test version or are there actually no dedicated 5 and 0 keys?

[jacobolus]

We’re talking about this one:

Is this a layout-only test version or are there actually no dedicated 5 and 0 keys?

Actually there aren’t any numbers at all (those are arrows and navigation keys, etc.), it’s a 40%-style layout. Numbers would be on a separate layer, either in numpad layout or something similar.

Anyway though, this discussion is a bit off topic for this thread. You might want to comment over at http://geekhack.org/index.php?topic=59396

[jacobolus]

interisting its like Rivertron

Show Image

Huh, I hadn’t seen that one before. I wonder if it ever got wired together.

[berserkfan]

Looking at you guys I feel so jealous. You have some of the best keyboards around, truly ergonomic and designed for your specs. Looks like I really have no choice but to work on my own customs.